Re: [VirtualGL-Users] Can a custom build of mesa and TurboVNC support 3D applications?

[DRC <dco...@us...>]

The way it works with VirtualBox is that you actually run the VM 
software in VirtualGL rather than running VirtualGL in the VM.  VBox 
already has its own mechanism for marshalling OpenGL calls from the 
guest to the host, so the host application works just like any other GLX 
app as far as VirtualGL is concerned.

KVM and Xen are a fundamentally different type of solution, though. 
They aren't self-contained and thus couldn't be run as applications in 
VirtualGL.  And when you try to run VirtualGL inside of the VM, it's 
going to fail unless the VM is virtualizing the GPU somehow.  If the VM 
can virtualize the GPU, then you really don't need VirtualGL, because 3D 
rendering should "just work".  If the VM can virtualize the OpenGL 
stack, as VirtualBox is doing, then you would need to run the VM 
software in VirtualGL as we're doing with VBox.

Note that it might be possible to open a display connection to :0.0 on 
the host and use VirtualGL inside the VM but set VGL_DISPLAY to point to 
the host display.  We don't generally support this configuration because 
of security issues (you would have to open up TCP connections on the 
3D-accelerated display, which is never a good idea in a production 
environment), but for a proof of concept, it might be OK -- assuming it 
works.

On 3/8/13 4:41 PM, James Wettenhall wrote:
> DRC,
>
>> Why can't you use VirtualGL?  The Chimera page clearly says that they
>> recommend "vendor-supplied graphics drivers", which VirtualGL would
>> enable you to use.
>
> A litte more information.
>
> Our CentOS 6 VMs have the xorg-x11-drv-qxl driver installed.
>
> I just upgraded to the latest version:
> https://rhn.redhat.com/errata/RHSA-2013-0218.html
>
> But trying to run glxgears with VirtualGL gives:
>
> Error: couldn't get an RGB, Double-buffered visual
>
> It may be that something needs to be tweaked at the hypervisor level:
>
> http://www.linux-kvm.org/page/SPICE
>
> I'll contact our KVM admins.